Yifan Mei

This paper presents a cautiously optimistic view of Large Language Model (LLM)-driven grief support technologies and their impact on user’s wellbeing. We focus on “Griefbot” technology: a type of chatbot designed to help individuals cope with the loss of loved ones by simulating interactions with the deceased. Drawing from cognitive science, we argue that grief is best understood as a domain-specific psychological mechanism. This mechanism is activated by the loss of significant relationships due to death. Grief process aims at adaptation and self-identity reconstruction. We contend that griefbots do not disrupt the grieving process but instead align with it, serving as tools that enhance individuals’ ability to navigate personal loss and reconstruct meaning. By situating Griefbot within the historical context of grief technologies—from Bronze Age artifacts to contemporary digital memorials—we argue that such LLMdriven tools complement, rather than replace, traditional grieving practices by offering a personalised and private space for reflection. Finally, we critically assess ethical concerns, including emotional dependency, autonomy, and the commodification of grief. We conclude that, if responsibly developed, Griefbot and similar AI technologies can enrich the grieving practices without fundamentally altering grief’s psychological underpinnings.

Several scientists and philosophers have argued that knowledge is more basic than belief, partly by appeal to evidence that non‐human primates attribute the former but not the latter. We re‐examine this evidence and argue that it is, in fact, more consistent with the presence of belief‐attribution than its absence. We also highlight the benefits of belief‐attribution and cast doubt on its alleged costs. While further research is needed to settle the debate, we recommend an attitude of cautious optimism: The understanding of the mind our close relatives possess may be more like ours than previously assumed.